Talk given by Gaute Einevoll, Norwegian University of Life Sciences.
Given to the Redwood Center for Theoretical Neuroscience at UC Berkeley on March 3, 2010.
Abstract. Mathematical modelling endeavours rely on experimental data to make progress, both to constrain and to test the models. For cortical neural network models the dominant experimental method in vivo has so far been single-unit extracellular recordings: when a sharp electrode is placed sufficiently close to the soma of a particular neuron, the recorded potential reliably measures the firing of individual action potentials in this neuron. This information is contained in the high-frequency part of the recorded potentials. The low-frequency part, that is, the local field potentials (LFP), has proved much more difficult to interpret and has typically been discarded.
Other experimental methods, particularly methods that measure population-level activity in vivo, are needed to facilitate development of biologically relevant cortical network models. Large-scale electrical recordings using various types of multielectrodes, i.e., electrodes with many contacts, are one such option. As techniques for such recordings are rapidly improving, there is a need for new methods for extraction of relevant information from such data.
Extracellular potentials in the brain are in general due to complicated weighted sums of contributions from transmembrane currents, and the potentials can be calculated by a combination of compartmental modelling providing the transmembrane currents following neural activity and electrostatic forward modelling using the quasistatic version of Maxwell’s equations. In the seminar, results from several projects aimed at elucidating the link between recorded extracellular potentials and the underlying neural activity, as well as extraction of neural network dynamics directly from multielectrode recordings, will be presented:
(1) investigation of how neural morphology and electrical parameters affect the shape and size of extracellular action potentials; (2) investigation of how the LFP generated by neurons in a population depend on synaptic activity, neuronal morphologies and frequency content; (3) introduction of laminar population analysis (LPA) where stimulus-evoked laminar-electrode data from rat barrel cortex are analysed in a scheme where the MUA and LFP are jointly modelled using physiological constraints; (4) extraction of thalamocortical and intracortical network models based on laminar-electrode data from barrel cortex and simultaneous recording of thalamic firing activity recorded in the homologous barreloid.